Characteristic Hypersurfaces and Constraint Theory Thesis submitted for the Degree of MSc by Patrik Omland Under the supervision of Prof. Stefan Hofmann Arnold Sommerfeld Center for Theoretical Physics Chair on Cosmology Abstract In this thesis I investigate the occurrence of additional constraints in a field theory, when formulated in characteristic coordinates. More specifically, the following setup is considered: Given the Lagrangian of a field theory, I formulate the associated (instantaneous) Hamiltonian problem on a characteristic hypersurface (w.r.t. the Euler-Lagrange equations) and find that there exist new constraints. I then present conditions under which these constraints lead to symplectic submanifolds of phase space. Symplecticity is desirable, because it renders Hamiltonian vector fields well-defined. The upshot is that symplecticity comes down to analytic rather than algebraic conditions. Acknowledgements After five years of study, there are many people I feel very much indebted to. Foremost, without the continuous support of my parents and grandparents, sister and aunt, for what is by now a quarter of a century I would not be writing these lines at all. Had Anja not been there to show me how to get to my first lecture (and in fact all subsequent ones), lord knows where I would have ended up. It was through Prof. Cieliebaks inspiring lectures and help that I did end up in the TMP program. Thank you, Robert, for providing peculiar students with an environment, where they could forget about life for a while and collectively find their limits, respectively. In particular, I would like to thank the TMP lonely island faction. It is thanks to Stefan Hofmann’s open mindedness and enthusiasm that I will attempt a part-time PhD. It seems truly special in the theoretical sciences that professors directly respond to and communicate with students. And so it was also, when Mark J. Gotay wrote a detailed list of master thesis proposals to an unacquainted student that had contacted him via email. My very special thanks goes out to Tine, for sharing all bitter times, albeit not being able to share the sweet times, of my bittersweet relationship with mathematical physics. Thank You. This thesis was written in LaTeX using MiKTeX and Emacs. Heavy use was made of the AMS-LaTeX packages and one or the other TikZ-picture shows up in this work. Thanks to all developers who make this available through some public license. Char. Hypersurfaces and Constraint Theory Contents 1. Introduction1 I. Mathematical Preliminaries4 2. Implicit Systems of Partial Differential Equations4 3. The Characteristic Initial Value Problem5 3.1. Basic Notions...................................5 3.2. The Wave Equation................................6 3.3. The more general Problem............................7 4. Characteristics9 5. Banach Spaces and Differential Calculus 11 5.1. Basics....................................... 11 5.2. The Functional Derivative and the Calculus of Variations............ 13 5.3. Symplectic Geometry............................... 14 5.4. Geometric Constraint Theory........................... 18 5.5. The Constraint Algorithm............................. 19 6. A Hamiltonian formulation of Classical Fields 22 6.1. The Jet Bundle.................................. 22 6.2. The Dual Jet Bundle............................... 23 6.3. Lagrangian Dynamics............................... 25 6.4. Cauchy surfaces and Space of Sections...................... 28 ∗ 6.5. Canonical Forms on T Yτ and Zτ ........................ 29 ∗ 6.6. Reduction of Zτ to T Yτ ............................. 30 6.7. Initial Value Analysis of Field Theories..................... 31 6.8. The Instantaneous Legendre Transform..................... 35 6.9. Hamiltonian Dynamics.............................. 37 6.10. Constraint Theory................................. 41 II. Thesis 42 7. Warm up: Parametrization- Invariant Theories 42 c Char. Hypersurfaces and Constraint Theory 8. Characteristic Hypersurfaces and Constraint Theory 44 8.1. Introduction.................................... 45 8.2. The Characteristic Hypersurface(s)........................ 46 8.3. The Constraints.................................. 47 8.4. Symplecticity of the Constraint Submanifold................... 51 8.5. Vacuum Electrodynamics on Minkowski (−1; 1; 1; 1):.............. 54 µν p µν 8.6. FAB / −gg with Lorentz Metric g ...................... 55 8.7. EM on curved spacetime............................. 57 8.8. More General................................... 60 8.8.1. Charged Scalar Field........................... 61 9. Conclusion 61 A. Appendix 62 A.1. Klein Gordon on a Light Cone.......................... 62 A.2. Vacuum Maxwell on a null hypersurface..................... 67 d Char. Hypersurfaces and Constraint Theory Introduction 1. Introduction In classical mechanics, at some point every physicist-to-be is asked to calculate the motion of a pendulum in three dimensions. The motion of the pendulum’s point-mass is easily seen to be constrained to lie within a sphere of radius the length of the string. Usually, one then switches to a spherical coordinate system with fixed radius and has the constraint thus implemented from the start. We will briefly recall what happens when continuing in Cartesian coordinates. [ST95] Take a free particle confined to the two-dimensional unit sphere in three-dimensional ambient space. Using a Lagrange multiplier λ and coordinates qi, i = 1; 2; 3, the Lagrangian reads m L(qi; q˙i; λ, λ˙) = q˙2 + λ(q2 − 1) 2 The Euler-Lagrange equations are of course mq¨i − 2λqi = 0 (evolution equation) q2 − 1 = 0 (constraint) Trivially, the constraint equation has consequences on the initial position we may choose: It has to lie on the sphere! But this is not the whole story. Quite formally, the whole system of equations reads mq¨i − 2λqi = 0 λ¨ = 0 qq˙ = 0 λ˙ = 0 q2 − 1 = 0 mq˙2 + 2λ = 0 In particular, the initial velocity needs to be tangent to the sphere. The Hamiltonian formalism lives on the space of initial conditions - so let’s see what happens, when we switch to phase space. For the canonically conjugate momenta pi, π, we calculate pi = mq˙i, π = 0. Hence, from the start phase space is constrained to 8 7 P = f(qi; λ, pi; π) 2 R j π = 0g R π = 0 is called a primary constraint and P the primary constraint submanifold of phase space. Primary constraints arise, when the Legendre transformation is not onto. The Hamiltonian becomes H : P −! R 1 (q ; λ, p ; π) 7−! p2 − λ(q2 − 1) i i 2m Note, that being odd-dimensional, P is not a symplectic submanifold w.r.t. the canonical symplectic form. The Hamiltonian is defined only on P, so we better restrict to initial values 1 Char. Hypersurfaces and Constraint Theory Introduction such that we do not flow off P by introducing new constraints: In view of more intricate systems, Dirac1 has developed an algorithm based on Poisson brackets to tackle this problem. In subsection 5.4 we introduce an algorithm due to Gotay, Nester and Hinds [GNH78] that is based on the presymplectic structure of P. In this case, also by simply using the Hamilton equations, we find the secondary constraints q2 = 1 ; pq = 0 ; p2 = −2mλ Secondary constraints are those that come from Hamilton’s equation by these considerations. We call the submanifold satisfying all constraints the final constraint submanifold and denote it by C. Finally, by a classification due to Dirac, all these constraints are second class and this classification allows us to calculate the dynamical degrees of freedom via 1 DoF = D − # − · # Dim:conf:space 1stClass 2 2ndClass 1 = 4 − 0 − · 4 2 Which is equal to two - the dimension of the sphere. The classification will be explained in section 5.4. Constraints are ubiquitous in field theory. In fact, if a field theory features gauge invariance, it necessarily features constraints (Proposition 5.5.3). The furthest reaching attempt to characterize the relationship between constraints and gauge invariance that the author is aware of, has been made in [GM06]. From there we take the Hamiltonian formulation of a field theory, recapitulated in section 6, and point out where the “gauge $ constraints” correlation starts off. In particular, we give an example in section 7. So where do characteristics come in? Now, hyperbolic partial differential equations are pri- marily defined through the existence of associated characteristic hypersurfaces. For physicists, hyperbolic PDEs are formally defined as those PDEs, which feature unique solutions for an initial-value problem and whose solutions propagate with finite speed. In section 4 we will make the connection. What makes characteristic hypersurfaces special in the theory of constraints is that they tend to bring about new constraints. This was already noted by [Ste80], who investigated quantization on characteristic hypersurfaces. We shall skip all the functional analysis involved for now (see section 5) and take as an example the massless Klein-Gordon field on a 4-Minkowski background: Z 1 L(φ, @ φ) = ηµν@ φ∂ φ ; η Minkowski Metric µ 2 µ ν 1P.A.M. Dirac. Generalized Hamiltonian dynamics. Can. J. Math., 2:129-148, 1950 2 Char. Hypersurfaces and Constraint Theory Introduction Z 1 1 = (@ φ∂−φ − @ φ∂ φ) ; x = x ; n = 2; 3 ; x± := (x ± x ) 2 + n n n n 2 0 1 Now, in the first coordinate system we choose x0 as the evolution direction and fx0 = 0g as initial value hypersurface. In the second, we choose x+ as the evolution direction and correspondingly fx+ = 0g as initial value hypersurface. Latter, as we will see in section 3, is characteristic w.r.t. the Euler-Lagrange equations. We calculate the canonical momenta π0 and π+, respectively @L π0 := = @0φ @(@0φ) x0=0 x0=0 @L π+ := = @−φ @(@+φ) x+=0 x+=0 In the first case, it is clearly possible to substitute the canonical momentum π0 for the ve- locity @0φ on the hypersurface and obtain the unconstrained phase space P0 = f(φjx0=0; π0)g. In the second case, this is not possible. Rather, we obtain the constrained subspace P+ = f(φjx+=0; π+) j π+ − @−φjx+=0 = 0g, where again, π+ − @−φjx+=0 = 0 is referred to as a primary constraint.